I wasn’t able to attend today’s hearing in San Francisco on Anthropic’s motion for summary judgment on fair use in Bartz v. Anthropic. So, please take my assessment with a grain of sale. Also remember that it’s risky, if not foolish, to predict how a judge will rule based on an oral argument.
But, based on the live tweeting of Adam Eisgrau (who admittedly is from Progress Chamber, a pro-tech group favoring fair use), and based on the amazing reporting of Annelise Levy of Bloomberg Law, Anthropic has to be pleased with how the oral argument went.
First off, as Levy reports, Judge William Alsup disclosed to the parties which way he is leaning: “I’m inclined to say they did violate the Copyright Act but the subsequent uses were fair use. That’s kind of the way I’m leaning right now…. Sometimes I say that and change my mind.”
Factor 1: Purpose and Character of Defendant’s Use
Just as with Judge Chhabria, Judge Alsup seemed to be leaning to recognizing the AI training as a transformative use under Factor 1. His questions about whether the Ninth Circuit had applied Authors Guild v. Google (to which Anthropic’s attorney replied that the Supreme Court had cited it 4 times in the Warhol decision) suggest that Judge Alsup is inclined to reach the same conclusion on Factor 1 as Judge Chhabria indicated he was favoring: the use of copyrighted works to train AI models is transformative. Judge Chhabria even said it might be called “highly transformative.”
Two other signs that suggest that Judge Alsup is leaning this way of finding a transformative purpose in AI training: (1) at one point he asked Anthropic’s attorney to move on to Factors 2 through 4 (which I often interpret as a good sign for the lawyer); and (2) his questions in response to the Bartz attorney’s attempt to characterize the use by Anthropic as the same as Napster (at least based on Easgrau’s account).
Factor 4: Potential Market Harm to Plaintiffs’ Works
The parties disagreed over whether lost licensing was a market harm under Factor 4. But, from the reporting above, there doesn’t appear to be much discussion of the novel theory of “market obliteration” that Judge Chhabria asked about at the hearing in Meta. According to the Bloomberg Law article, Judge Alsup was skeptical there was market harm for plaintiffs’ books given the “guardrails Anthropic has in place to prevent verbatim outputs.” And, when the plaintiffs’ lawyer replied that “AI can dilute the market by flooding it with cheap knockoffs,” the Judge was “unsure about the harms of dilution, pointing out common similarities in mystery novels involving murder.”
There also doesn’t appear to be any questions about the Copyright Office’s controversial pre-publication report, which was cited by Bartz as supplemental authority (because it was issued after the briefing was all done).
Pirated books datasets
But it’s worth noting: Judge Alsup did express concern about Anthropic’s use of pirated books datasets–as did Judge Chhabria at the fair use hearing in Kadrey v. Meta. Judge Alsup questioned Anthropic’s attorney: “I have a hard time seeing that you can commit what is ordinarily a crime, but get exonerated because you end up using it for a transformative use.” Judge Chhabria expressed a similar concern, suggesting that allowing it seemed “messed up.”
Judge Alsup asked for supplemental briefing on whether Google’s copying of the books for its database for Google Books search was considered fair use under Authors Guild v. Google.
Related Stories